Summary of Server and Network Equipment Products
Servers
SSD
- Servers typically support up to 8-12 3.5-inch hard drives or 24 2.5-inch hard drives, depending on customer requirements and server capabilities.
U.2
Enterprise-grade
- High performance, with extremely fast read and write speeds
- Supports multiple queues and higher IOPS
- Relatively high cost
- Popularity needs to be further improved
SAS
Enterprise-grade
- Good compatibility, downward compatible with SATA
- Suitable for enterprise-level applications, high stability
- Higher price
- Relatively complex equipment and cables
M.2
Consumer-grade
- Compact size, does not occupy too much space, not often chosen for server platforms
- Fast transfer speeds, especially M.2 with NVMe support performs exceptionally well
- Potential for heating issues, requiring good heat dissipation measures
- Relatively limited capacity: While capacity is continuously increasing, high-capacity options can be costly compared to some traditional hard drives
SATA
Mass-market grade
- Mature technology, low price
Compatibility The devices have broad compatibility, supporting a wide range of systems and applications.
Transmission Speed The transmission speed is relatively slower, gradually unable to meet the demands of high-performance applications.
Size The physical size is relatively large, corresponding to 3.5-inch drive bays.
CPU
- Servers typically adopt a dual-processor design, supporting two CPUs of the same platform.
- Intel
- Platinum
- The Platinum series often offers the highest performance, suitable for tasks requiring extreme computational power, such as advanced analytics, machine learning, and AI workloads.
- Gold
- The Gold series provides improved 4-socket performance, built-in workload acceleration, and advanced security technologies, optimized for demanding mainstream data centers, cloud computing, networking, and storage workloads.
- Silver
- The Silver series delivers essential performance and efficiency for entry-level data center computing, networking, and storage.
- Bronze
- The Bronze series offers hardware-enhanced reliability, availability, and serviceability for small businesses and basic storage servers.
- EPYC (AMD's Ryzen Threadripper for servers)
- Core Count
- Different EPYC processor models have varying core counts, ranging from 8, 16, 32, to 64 cores.
- Thread Count
- The thread count typically corresponds to the core count, for instance, an 8-core processor may have 16 threads.
- Base Clock and Turbo Boost
The base clock speed is the fundamental operating frequency of a processor, while Turbo Boost is the automatically elevated frequency when higher performance is required. Processors of different models have varying base clock and Turbo Boost speeds.
DDR
- Also known as DIMM, typical platforms offer 24 slots, with each slot supporting up to 128GB, depending on the server's parameter configuration.
DDR4
- The most common type of RAM used in servers, typically with frequencies of 2133, 2400, 2666, and others. The maximum supported frequency depends on the server platform.
DDR5
- The next-generation memory standard, primarily used in Intel's second-generation and later server platform architectures. DDR5 offers higher bandwidth and transfer rates, with starting frequencies typically at 4800MHz and above.
SLDDR
- Advantages: Low-power RAM suitable for specific customers.
- Disadvantages: Slower data transfer speed and relatively lower performance.
ECC
- Advantages: Provides error checking and correction functionality, enhancing data reliability, suitable for applications requiring high data accuracy such as servers and workstations.
- Disadvantages: Relatively higher price, approximately 20% more expensive than regular RAM.
REG
- Advantages: Adds registers on top of ECC, improving memory stability and compatibility, supporting larger capacities.
- Disadvantages: Price is about twice as high as regular RAM, with slightly higher latency, affecting performance slightly.
RECC
- Advantages: An abbreviation for REG ECC, sharing similar advantages as REG ECC, including error checking and correction functionality, while supporting larger capacities.
- Disadvantages: Higher price, slightly higher latency, requiring support from server motherboards and may not be usable on regular motherboards.
Here is the English translation of the provided server hardware-related text about GPUs, motherboards, and network cards:
GPUs
GPUs typically require newer platform architectures in servers to fully utilize their computational advantages. Such platforms are often in the 4-6U size range.
Nvidia-A100
-
- Features high bandwidth, high computational power, and low latency, suitable for artificial intelligence, deep learning, scientific computing, and more.
- Comes in 40G customized versions and 80G PCIE factory versions, with significant price fluctuations.
-
Nvidia-A800
-
- A replacement product for the A100, mainly reducing the NVLink transfer rate from 600GB/s of the A100 to 400GB/s, while other parameters remain largely the same.
- A special version with slightly reduced performance compared to the A100.
-
A40, A30, L20, L4, etc.
-
- Alternative versions with lower prices, depending on the customer's configuration scheme and needs.
- Suitable for graphics processing, virtual reality, film post-production, inference, deep learning training, and high-performance computing.
- Consideration should be given based on specific requirements and application scenarios.
-
Nvidia-RTX4090
-
- Boasts extremely high computational and graphics processing capabilities, capable of handling complex computational tasks and graphics rendering needs.
- Can be used for some AI-related tasks, such as deep learning inference.
- Not specifically optimized for servers: Compared to dedicated server-grade GPUs, it may have slightly inferior adaptability and stability in certain server-specific scenarios.
Motherboards
- Each generation of equipment platforms is built on an entirely new platform architecture based on the motherboard, and the price of a new platform is often very high.
Network Cards
When it comes to server hardware, there are several key components and features to consider:
Network Interface Cards (NICs): Commonly available in dual-port, quad-port, or other configurations, NICs can offer 4×1GE electrical ports or 2×10GE electrical/optical ports, depending on the customer's requirements.
Cooling: Modular fans are typically used for cooling, and for high-end computing equipment, immersion cooling solutions may be employed to effectively dissipate heat.
RAID Cards: Also known as array cards, RAID cards primarily serve to protect data. In certain RAID levels, such as RAID 1 and RAID 5, even if some disks fail, the integrity and availability of data are still guaranteed.
Power Supplies: Common 2U rack-mounted servers often feature 1+1 modular power supply units that allow for hot-swapping. Power supplies come in different ratings, including Titanium and Platinum.
- Titanium: High-efficiency power conversion, suitable for high-performance servers requiring stable power output of 2000w, 3000w, and more.
- Platinum: Commonly rated below 1100W, with modular versions available in 550w, 750w, and other ratings. These also support hot-swapping.
Rails: Sliding rails, also known as server rails, are used to securely mount servers within a rack.
Core Switches: Core switches are crucial devices in a network architecture, typically operating at the third layer. One or two core switches are commonly deployed per rack.
- Functionality: Core switches support VLAN partitioning, link aggregation, routing protocols, and other features to enable fast communication between different network segments and ensure stable network operation.
- Size: Core switches typically have a large number of ports, with common configurations including 24-port (1-2U) and 48-port (2-3U) models.
- Ports: Optical ports are also available on core switches, providing high-speed, long-distance connectivity options.
Network Interface Cards (NICs):
10GE (10 Gigabit Ethernet) and 40GE (400 Gigabit Ethernet):
-
- Optical modules such as XFP, QSFP, and QSFP+ are used to enable these high-speed connections.
-
Electrical Ports:
-
- For 1GE (Gigabit Ethernet) connections, optical modules like SFP and SFP+ are common.
Brands:
- Popular brands include Cisco, H3C, Ruijie, and Juniper, among others, offering various NIC solutions.
Core Routers:
The central device in network architectures, core routers are expensive and critical to network operations.
Expansion Modules (or Line Cards):
-
- Core routers often have multiple MPC (Modular Port Concentrator) expansion modules to provide ports like 100GE to 400GE.
-
Ports:
-
- Core routers can offer a wide range of ports from 1GE to 400GE, often with hundreds of ports available.
-
Size:
-
- Enterprise-grade core routers typically have a size ranging from 6-8U and a weight of around 80KG.
-
Modules:
SFP:
Commonly used in Gigabit Ethernet and Fiber Channel networks for short- to mid-range transmissions, such as SFP GE electrical modules (100m, RJ45).
XFP:
Supports up to 10Gbps, mainly used in 10GE, SONET/SDH, and other high-speed networks for mid- to long-range transmissions, such as 10G XFP dual-fiber optical modules with a wavelength of 850nm.
QSFP:
-
- Offering speeds like 40Gbps and 100Gbps, QSFP modules are primarily used in data centers, high-performance computing, and other applications requiring short- to mid-range high-speed transmissions, like QSFP+ 40G optical modules (850nm, 100m, SR4, supporting 1-to-4 breakout).
o Single-mode
§ Long transmission distance: It can achieve long-distance transmission.
o Multi-mode
§ Short-distance transmission: Usually used for the internal lines in the machine room building.
· Board card
o An expansion module based on the core router
o There are multiple expansion modes
o A board card is required to achieve the ultra-high network transmission efficiency of the core router
o Depending on the size of the port, the price also ranges from several thousand to several hundred thousand.
Seagate Exos X18 ST16000NM000J 16TB 3.5 LFF 6Gbps 7.2K RPM 512e Enterprise SATA Hard Drive - Brand New Seagate Exos X18 ST16000NM000J 16TB 3.5 LFF 6Gbps 7.2K RPM FastFormat 512e Enterprise SATA Hard Drive - Brand New
EXOS X18
Seagate manufactures hard drives that specifically address the needs of the hyperscale storage market. As the flagship of the Seagate® X class, the Exos™ X18 enterprise hard drives are the highest-capacity hard drives in the fleet.
Best-Fit Applications
- Hyperscale applications/cloud data centers
- Massive scale-out data centers
- Big data applications
- High-capacity density RAID storage
- Mainstream enterprise external storage arrays
- Distributed file systems, including Hadoop and Ceph Enterprise backup and restore- D2D, virtual tape
- Centralized surveillance
• All 4Kn/512e drives ship in 512-byte emulation mode but can be
easily reformatted to 4Kn (and vice versa) with the FastFormat
feature.
• FastFormat only works with 512e and 4Kn. You can't use FastFormat with 512-byte native drives.
• FastFormat reformats the drive, so all data on the drive will be lost once you issue the command. Don't forget to back up your data before issuing FastFormat!
• T10 fast format conversion between 4K and 512E configurations is possible only if sector sizes are exact multiples of 8.
• The new sector size will take effect only after a fast format.
Compatibility: Compatible with systems that support 16TB 3.5" SATA HDDS
SSD: U.2/SATA/SAS/M.2 CPU: Platinum/Gold/Silver/Bronze/AMD DDR: DDR3/DDR4/DDR5/ECC GPU: H100/A100/H800/A800/4090/3090 Power supply: platinum / titanium / 750w / 550w / rail Core switch: VLAN/24/48/10GE/1GE/SFP Core router: MPC/40GE/100GE/400GE/board Module: SFP/XFP/QSFP/single mode/multimode/jumper Server: 1U / 2U / rack / storage / 730XD / 740XD Brand: Intel/Xeon/EPYC//Cisco//H3C/Juniper
Audited Supplier 
){kind=link}